1820 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 52, NO.
8, AUGUST 2004
Mixed-Potential Volume Integral-Equation
Approach for Circular Spiral Inductors
Behzad Rejaei
Abstract—The electromagnetic behavior of circular spiral in-
ductors on layered substrates is analyzed by using the concentric
ring approximation. The problem is formulated in terms of mixed-
potential volume integral equations. The latter are semianalytically
solved within the quasi-one-dimensional approximation. The result
is a fast computationally efficient model, which takes into account
substrate RF loss, as well as nonuniform current density distribu-
tion across inductor windings. The results are in good agreement
with experimental data from various inductors on both low- and
high-resistivity silicon substrates.
Index Terms—Eddy currents, mixed-potential integral-equation
(MPIE) method, monolithic inductors, spiral inductors, substrate
RF loss.
I. INTRODUCTION
T HE DEMAND for full integration of RF transceiver sys-
tems has been the driving force behind the recent surge
of interest in on-chip inductors and—to a lesser extent—trans-
formers. Emphasis lies on integration in silicon (Si) technology,
which offers the most in terms of reliability, low cost, and a high
level of integration. The major drawback of Si-based devices is,
however, their proneness to significant RF losses caused by the
current flow in the conductive substrate.
Although the inductance of lumped elements can be predicted
with reasonable accuracy [1]–[6], the analysis of substrate RF
losses is not straightforward. The common strategy is, therefore,
based on empirical or semiempirical lumped-element represen-
tations [7]–[11], where substrate effects are modeled by using
lumped capacitors and resistors, determined through compar-
ison with experimental data. Despite its inherent simplicity, this
method lacks the predictive power required for component de-
sign and optimization.
An alternative approach is to use comprehensive com-
putational techniques [12]–[19] or semianalytical methods
[20], [21]. In [12], inductor substrate losses were treated
using a full-wave electromagnetic (EM) numerical package.
The spectral-domain method of moments (MOM) was ex-
ploited in [13] to analyze spiral inductors on multilayered
Si substrates. Reference [17] presented an analysis of planar
[two-dimensional (2-D)] devices based on the mixed-potential
Manuscript received November 20, 2003; revised February 19, 2004.
The author is with the Faculty of Electrical Engineering, Mathematics, and
Computer Science, Laboratory of Electronic Components, Technology, and
Materials, Delft Institute of Microelectronics and Submicron Technology,
Delft University of Technology, 2628 CT Delft, The Netherlands (e-mail:
b.rejaei@ewi.tudelft.nl).
Digital Object Identifier 10.1109/TMTT.2004.831567
0018-9480/04$20.00 © 2004 IEEE
Authorized licensed use limited to: Tsinghua University. Downloaded on April 02,2023 at 11:57:47 UTC from IEEE Xplore. Restrictions apply.
integral-equation (MPIE) method, generalized to three-dimen-
sional (3-D) conductors in [18]. Reference [19] used the MOM
and the finite-difference time-domain technique to analyze the
nonuniform current distribution in inductor windings. The work
of [20] was based on a volume integral-equation approach,
but used the Greenhouse algorithm [2] to evaluate the partial
inductances. In [21], this method was generalized to include
substrate eddy currents, which are significant in highly doped
low-resistivity substrates.
In spite of their power and generality, these numerical tech-
niques usually do not offer the high computational speeds
required for inductor optimization in mixed-level circuit/com-
ponent design. This is because they require the division of the
conductor volume (or surface) into a relatively large number
of segments, and numerical evaluation of the corresponding
Green’s matrix elements on arbitrary substrates. Significant
reduction of computation time could be achieved if the field
equations were to be analytically solved in one direction (e.g.,
along the windings), effectively reducing the dimensionality
of the problem. Such a model was previously presented by
Rejaei et al. [22] by using the concentric-ring approximation
of circular spiral inductors [3]. Here, a planar circular spiral
inductor is replaced by a series connection of concentric rings,
and the resulting rotational symmetry is exploited to semiana-
lytically solve the field equations within the MPIE formalism.
It should be mentioned that expressions for the inductance
and quality factor of a single ring above a layered (magnetic)
substrate were already derived in [23], but did not account
for losses due to capacitive coil-substrate coupling.
In this paper, we present the full mathematical framework
of the method of [22] and extend the latter to 3-D conduc-
tors, including the effect of the nonuniform current distribution
(skin effect, current crowding) in the coil windings. Employing
the MPIE technique, a solution is obtained by using a discrete
Fourier transform along the rings, followed by application of
Galerkin’s method to compute the field distribution across each
ring. The result is a fast predictive model for circular spiral in-
ductors on layered substrates. The model is derived in Section II
and compared to measurement data in Section III.
II. CONCENTRIC-RING MODEL
The concentric-ring model of circular spiral inductors [3] re-
sults from replacing an -turn spiral by a series connection of
circular rings (Fig. 1). It is assumed that the width and
spacing of the rings are identical to those of the spiral in-
ductor. The overall dimension of the concentric-ring model is
REJAEI: MIXED-POTENTIAL VOLUME INTEGRAL-EQUATION APPROACH FOR CIRCULAR SPIRAL INDUCTORS 1821

Fig. 3. Cross section of the system of rings. The volume and vertical cross
i V S s
section of the th ring are denoted by and , respectively. is the unit normal
Fig. 1. Concentric-ring model of a circular spiral inductor. The width andw to the conductor surface.
s
spacing of the rings equal those of the inductor. The radius R is determined
by requiring the outermost ring to have the same length as the first winding of
the spiral inductor. where is the conductor resistivity, is the electric field, and
and are the vector scalar potentials, respectively. The latter
can be expressed as

(2)

(3)

Fig. 2. N
System of concentric rings as a 2 -port system. where is the total volume of the conductor, de-
notes the charge density, , , and
determined by equating the total lengths of the inductors (at the . The form of the 3 3 matrix and the
midline).1 scalar Green’s functions depend on the gauge chosen [16].
To fully exploit the rotational symmetry of the model, we next We choose for the representation in [14] where
assume that the rings are disconnected. The set of concentric
rings can then be viewed as a -port system with inputs (4)
and outputs (Fig. 2). Once the admittance matrix (or, equiva-
lently, the impedance matrix) of the -port system is calcu- Note that, in (2) and (3), the dependence on the Green’s func-
lated, the admittance matrix of the spiral inductor is retrieved tions on frequency is not shown explicitly.
by connecting the rings again. Note that, since the calculation Using (1)–(3) and the continuity equation ,
of the admittance matrix of the system of rings is independent one arrives at the closed set of integro-differential equations
of the way the terminals are interconnected, this method can
be easily generalized to investigate various structures such as (5)
transformers and stacked inductors built on several metal layers
over layered substrates. In the remainder of this section, we will
present the derivation of the model in detail in several steps. (6)

A. MPIEs
The starting point of our analysis is the MPIE approach
[14]–[18]. Consider a multiply connected conductor consisting
of concentric rings with the volumes , , the conductor ( is the unit vector normal to the surface). On the
placed on a substrate consisting of infinitely extended planar
layers parallel to the – -plane (Fig. 3).
Neglecting the Lorentz force, the current density at each
point inside the conductor is given by the Ohm’s law
(1)
In order to allow a complete solution, (5) and (6) should be
supplemented by boundary conditions for and/or the normal
component of the current density on the surface of
interfaces between external contacts and the conductor, either
or should be specified, whereas should vanish everywhere
else on the surface of the conductor.
In this paper, we shall restrict ourselves to substrates con-
sisting of isotropic layers, i.e., where each layer is described by
isotropic constitutive parameters. The resulting rotational sym-
metry of the system is then reflected in the relations

1Simulations carried out by using HP ADS/Momentum (not shown here)

(7a)
demonstrate that the concentric ring approximation is fairly accurate even if
the shape of the inductor windings significantly deviates from that of an ideal (7b)
ring, e.g., whens>w . The resulting error in inductance and quality factor was
found to be less than 4% for typical on-chip inductors built on lossy substrates.
Furthermore, many practical structures are not spirals, but symmetric inductors The spectral-domain representation of the Green’s functions
where the actual windings are indeed rings. and is given in the Appendix.

Authorized licensed use limited to: Tsinghua University. Downloaded on April 02,2023 at 11:57:47 UTC from IEEE Xplore. Restrictions apply.
1822 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 52, NO. 8, AUGUST 2004

Equations (12a)–(12d) can be used to transform (9) and

(10) into a set of decoupled integral equations by applying the
Fourier transformations

(13a)

Fig. 4. Concentric-ring system in cylindrical coordinates. u is the azimuthal

!
(13b)
unit vector. The input and output of the rings are denotes by the limits  0
and  ! 2 .

(13c)
B. Quasi-One-Dimensional (1-D) Approximation
Employing cylindrical coordinates , we next adopt the
quasi-1-D approximation by neglecting the components of (13d)
and variations of in the radial - and vertical -directions
inside the rings
After multiplying (9) and (10) by , calculating the integral
over from 0 to , and using (13a) and (13b), one arrives at
for (8)

where is the azimuthal unit vector (Fig. 4). Here, in fact, we

have assumed that current follows the direction of the rings in-
side the conductor. Using (5)–(8), one then arrives at the integral (14)
equations

(15)
(9)
Here, the integrations are carried out over the cross sections
of the rings in the – -plane (perpendicular to the azimuthal unit
(10) vector ), , and

(16a)
where and
(16b)
(11)
(16c)
In (9) and (10), and are periodic func-
tions of with the period . This is because
in cylindrical coordinates. One can, (16d)
therefore, expand and in the Fourier series

(12a) (16e)

(12b) where , and , denote the potential

and current density on the input and output
cross sections of the th ring, respectively (Fig. 4). Note that
(12c) and are constant on , but depend on the ring index .

C. Solution of the Integral Equations

(12d) For each harmonic mode , (14) is a Fredholm integral
equation of the second kind for the functions , whereas
(15) is a Fredholm integral equation of the first kind for the
Note that and because and actu- functions . Both equations can be numerically solved
ally are even functions of due to reflection symmetry in the by Galerkin’s method, where one approximates and
– -plane. by an expansion in terms of a finite number

Authorized licensed use limited to: Tsinghua University. Downloaded on April 02,2023 at 11:57:47 UTC from IEEE Xplore. Restrictions apply.
REJAEI: MIXED-POTENTIAL VOLUME INTEGRAL-EQUATION APPROACH FOR CIRCULAR SPIRAL INDUCTORS 1823

of basis functions , respectively, on as

follows:

(17a)
Fig. 5. Vertical cross section of each ring (in the r –z -plane) is divided into
rectangular segments.

(17b)
D. Admittance Matrix of the System of Rings
Multiplication of both sides of (14) and (15) by the basis func- In order to calculate the admittance matrix of the concen-
tions , respectively, , and integration over the ring cross tric-ring system, one has to find the relationship between the
sections results in matrix equations for the constants and ring potentials and the ring currents
with the solution
(22)

(18)
at the inputs and outputs . Substitution of
the Fourier series expansion (13c) in the above equation results
(19) in

(23a)
where

(23b)
(20a)

where are the Fourier components of . By integrating (17a)

and (17b) over , and using (16a)–(16e), (18), (19), (20d), and
(22)–(23b), one obtains the matrix equations
(20b)
(24)

(20c)
(25)

in which
(20d)

(26a)
with denoting the Kronecker delta.
In order to reduce the computation time for the integrations
in (20b) and (20c), we divide the cross section of the rings in (26b)
rectangular segments and use the basis functions
are matrices, and
if
otherwise (27)
if
(21) designate the total current at the input, respectively, and output
otherwise
of the th ring. (The minus sign has been included according to
where and denotes the surface of the th the conventional definition of port currents in network theory.)
rectangular segment on the cross section of the th ring (Fig. 5). From (24) and (25), one can determine the Fourier coefficients
In the Appendix, we show that each element of the matrices and in terms of and . After sub-
and can then be calculated by performing a single stituting the solution obtained (not shown here) in the Fourier
integration in the spectral domain. series (13d) and (23b), evaluating and in the limits

Authorized licensed use limited to: Tsinghua University. Downloaded on April 02,2023 at 11:57:47 UTC from IEEE Xplore. Restrictions apply.
1824 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 52, NO. 8, AUGUST 2004

and , and applying several matrix manipulations,

we arrive at the admittance matrix as follows:

(28a)

(28b)

(28c)
(28d)

where and . Note that the summation in (28b)

only involves nonnegative ’s. This is because, as mentioned
previously, and which, in turn, im-
plies that and . Equations (28a)–(28d)
Z Y
Fig. 6. Decomposition of the matrices  and  in terms of mutual
together with the definitions (26a) and (26b) express the admit- resistance R , inductance L
 , conductance G
 , and capacitance C
tance matrix of the concentric-ring system in terms of the matrix matrices for each mode m.
elements of the Green’s functions and .
It is important to note that and have, in fact, been
computed from the vector-potential Green’s function and the
scalar potential Green’s function , respectively. It is, therefore,
natural to link and to the inductance, respectively, ca-
pacitance matrix of the system in the th mode by introducing
the decompositions

(29)

in which , , , and are real frequency-dependent

matrices (see Fig. 6). Here, can be interpreted as
the inductance matrix of mutually coupled inductors. Simi-
larly, represents mutually coupled capacitors. The ma-
trix takes account of the dc loss in the conductor, as well
as additional RF loss due to nonuniform distribution of current
density on the cross section of the windings (skin effect, cur- Fig. 7. Admittance matrix of the inductor is obtained by connecting the rings
in series. The underpass is modeled as a stripline, capacitively interacting with
rent crowding). However, it also includes losses due to substrate 0
the inductor windings through the N 1 metal–oxide–metal capacitors formed
eddy currents induced by the RF magnetic field. The matrix between the inductor and underpass metals.
, on the other hand, contains substrate RF losses caused by
the flow of substrate currents induced by the potential differ-
ence between different points on the conductor and between the TABLE I
conductor and reference ground. The latter has implicitly been LAYER PARAMETERS FOR MEDIUM-RESISTIVITY (MR)
AND LOW-RESISTIVITY (LR) Si SUBSTRATES
defined as a point at the infinity, where the Green’s functions
vanish.
In practice, only a small number of modes (small ’s) will
suffice to provide an adequate solution of the problem. Higher
modes (larger ’s) become important at very high frequencies
where the effective wavelength becomes smaller than the length
of rings. The number of modes included in the calculation de-
pends on the required accuracy and should be determined from
numerical experimentation.
2 2 admittance matrix of the spiral inductor. In practical
E. Retrieving the Admittance Matrix of the Inductor applications, however, an underpass is used to connect the inner
After the admittance matrix of the concentric-ring system terminal of the coil to other devices. It is not possible to directly
is computed, the rings can be connected in series to obtain the include the underpass in the concentric-ring model due to the

Authorized licensed use limited to: Tsinghua University. Downloaded on April 02,2023 at 11:57:47 UTC from IEEE Xplore. Restrictions apply.
REJAEI: MIXED-POTENTIAL VOLUME INTEGRAL-EQUATION APPROACH FOR CIRCULAR SPIRAL INDUCTORS 1825

Fig. 8. Measured (markers) and simulated (solid lines) inductance and quality factor of circular spiral inductors. (a) Inductor A (N = 4, R = 120 m, w =
13:7 m, s = 10:27 m). (b) Inductor B (N = 4, R = 140 m, w = 11:38 m, s = 3 m). (c) Inductor C (N = 6, R = 160 m, w = 11:25 m,
s = 3 m). Substrate resistivity was 5
cm. Metal thickness was 1 m.
1

breaking of the circular symmetry. Nevertheless, in the special as a stripline interacting with the inductor through lumped
case of Fig. 7, one can approximately model the underpass metal–oxide–metal capacitors formed between the underpass

Authorized licensed use limited to: Tsinghua University. Downloaded on April 02,2023 at 11:57:47 UTC from IEEE Xplore. Restrictions apply.
1826 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 52, NO. 8, AUGUST 2004

Fig. 9. Measured (markers) and simulated (solid lines) inductance and quality factor of circular spiral inductors. (a) Inductor A (N = 4, R = 120 m, w =
13:7 m, s = 10:27 m). (b) Inductor B (N = 4, R = 140 m, w = 11:38 m, s = 3 m). (c) Inductor C (N = 6, R = 160 m, w = 11:25 m,
s = 3 m). Substrate resistivity was 0.01
cm. Metal thickness was 4 m.
1

and windings. Despite the neglect of the mutual inductance III. EXPERIMENTAL RESULTS
between the underpass and spiral coil, this approximation turns
out to be accurate enough for typical integrated inductors, as We have fabricated various spiral inductors on medium-resis-
shown by the experimental data in Section III. tivity (5 cm) and low-resistivity (0.01 cm) Si substrates.

Authorized licensed use limited to: Tsinghua University. Downloaded on April 02,2023 at 11:57:47 UTC from IEEE Xplore. Restrictions apply.
REJAEI: MIXED-POTENTIAL VOLUME INTEGRAL-EQUATION APPROACH FOR CIRCULAR SPIRAL INDUCTORS 1827

TABLE II
GEOMETRICAL PARAMETERS (N , R, w , s), INDUCTANCE AT MAXIMUM QUALITY FACTOR (L ), MAXIMUM QUALITY FACTOR (Q ), AND
SELF-RESONANCE FREQUENCY (f ) OF THREE INDUCTORS ON MEDIUM- AND LOW-RESISTIVITY SUBSTRATES

Each substrate consists of a 525- m-thick Si layer, insulated

by a 4- m-thick layer of SiO . The spiral coils were built by
using the top metal layer (1- and 4- m-thick aluminum on 5-
and 0.01- cm substrates, respectively). The underpass con-
tact was built by using an additional 0.6- m-thick metal layer,
spaced 0.8 m below the top metal layer.
The -parameter measurements were performed using an
HP8510B/8514A network analyzer in combination with a Cas-
cade Microtech Summit 9000 probe station with 200- m-pitch
signal–ground probes. The calculated -parameters, after
deembedding the measurement pads, are then used to calculate
the inductance and quality factor of each coil in the
configuration where port 2 (connected to the inner terminal) is
grounded as follows:2

(30)
Table I lists the substrate parameters used in the simulations.
1 cm Si substrate using a 1-, 2-, and 4-m-thick Al layer.
The simulations were performed retaining the harmonic
mode only since numerical experimentation revealed negligible
improvement upon inclusion of higher modes. In order to take
account of the nonuniform distribution of current and charge
density, the cross section of each winding was divided into a
number of rectangular segments (see Section II-C and the Ap-
pendix). The simulations were carried out on an HP Unix work-
station. For a four-turn inductor, the computation time for each
frequency point was less than 0.22 s at five segments/winding
and less than 1.5 s at 25 segments/winding. The results shown
below were obtained using 25 segments/winding.
Figs. 8 and 9 show the measured and simulated and
for three different inductors (A,B, and C), on the medium-
and low-resistivity substrates. The geometrical parameters
of the three coils, as well as their measured and simulated
inductance at maximum quality factor , maximum
quality factor , and self-resonance frequency are
listed in Table II. The discrepancy between the measured and
simulated inductance curves for coils B and C in Fig. 8 can be
attributed to the underestimation of their by the model on
the 5- cm substrate (Table II). However, one should note that
the actual resistivity of the Si substrate may have differed from
2Unlike classical microwave substrates, Si technology does not allow
the use of a through-wafer vias providing access to a backside metal layer
(RF ground). In the experiments discussed here, large rectangular patches
(800 m 2 300 m) distanced 300 m away from the edge of the coils and
built in the first metal layer were used as a ground pad for the probes. The
large area of each patch provides a low-impedance path through the conductive
substrate to the metallic chuck on which the wafers are placed during the
measurement. The chuck replaces the backside metal as the RF ground.
Fig. 10. Measured (markers) and simulated (solid lines) quality factor of
inductor A (N = 4, R = 120 m, w = 13:7 m, s = 10:27 m) built on
the 5-
the nominal value used (5 cm). In conventional processes
variations up to 5% are normal on different wafers or different
locations on the same wafer. Besides affecting , this would
shift , in particular, on medium-resistivity substrates [24].
Nevertheless, the overall agreement between simulated and
experimental ’s and ’s on both substrates is good, bearing
in mind that no fitting parameters were used.
A common shortcoming of many commercial, 2-D and
two-and-one-half-dimensional (2.5-D) EM solvers is their
inability to handle the finite thickness of the conductors: the
metal layers are assumed to be infinitely thin, while the nonuni-
form current distribution in the vertical direction is accounted
for by using an effective complex frequency-dependent sheet
resistivity. The concentric-ring model, however, is not restricted
to this planar approximation. In order to test the accuracy of
the model in handling thick conductors, we built additional
spiral inductors on the 5- cm substrate with the thickness of
the top metal layer ranging from 2 to 4 m. The quality factor
of the 2-nH inductor built using a 1-, 2-, and 4- m-thick layer
is shown in Fig. 10. Again, a good match is found between
the simulated and measured data, although the model predicts
a slightly higher at frequencies above 6 GHz as the metal
thickness is increased.
IV. CONCLUSIONS
We have presented a semianalytical model for circular spiral
inductors on layered substrates. The model employs the concen-

Authorized licensed use limited to: Tsinghua University. Downloaded on April 02,2023 at 11:57:47 UTC from IEEE Xplore. Restrictions apply.
1828 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 52, NO. 8, AUGUST 2004

where and are substrate reflection coefficients deter-

mined by the recursive relations

(33a)

(33b)

(33c)

(33d)

(33e)

(33f)

where is the number of substrate layers (the infinite layer

Fig. 11. Multilayered substrate used for the evaluation of the spectral-domain beneath the substrate is denoted by ), and , , and
Green’s functions.
are the dielectric constant, permeability,
and conductivity of each layer, respectively (Fig. 11). The layer
tric-ring approximation where the circular inductor is replaced thicknesses are denoted by .
by the series connection of concentric rings. Using the rotational From (11), (12c), (12d), (31a), (31b), and Neumann’s addi-
symmetry of the system of rings, as well as the quasi-1-D tion theorem for Bessel functions [25]
approximation, a semianalytical solution is derived for the
MPIEs based on the Fourier series expansion of the potentials (34)
and currents in the rings. The results are in good agreement
with experimental data for inductors of various sizes and metal
thicknesses built on a lossy Si substrate. with , it follows that

APPENDIX
In order to calculate the functions and and their ma- (35a)
trix elements, we start with the spectral representation of the
Green’s functions and

(31a)

(31b)
where denotes the th-order Bessel function of the first kind
and and are spectral-domain Green’s functions.
Although the expressions for the spectral-domain Green’s
functions in stratified media can be found in various papers, we
present them below for the sake of completeness. We restrict
ourselves to the typical configuration of Fig. 11 where and
are both located above the substrate. (Although relevant, the
more general, but complicated case where and lie in arbi-
trary substrate layers will not be discussed here.) The functions
and are then given by
(35b)
The matrices and are evaluated by substituting
the above relationship into (20b) and (20c) and carrying out the
integrations using the basis functions (21). Due to the exponen-
tial dependence of the spectral-domain Greens functions on
and , the integrations over those variables can be performed
analytically, whereas those involving are calculated by means
of the relationships

(32a)

(32b) (36)

Authorized licensed use limited to: Tsinghua University. Downloaded on April 02,2023 at 11:57:47 UTC from IEEE Xplore. Restrictions apply.
REJAEI: MIXED-POTENTIAL VOLUME INTEGRAL-EQUATION APPROACH FOR CIRCULAR SPIRAL INDUCTORS
where and are Sturve functions. As a result, computation
of each matrix element only requires the evaluation of a single
integral over the variable .
ACKNOWLEDGMENT
The author would like to thank J. N. Burghartz, H. Schellevis,
and A. Akhnoukh, all of the Delft Institute of Microelectronics
and Submicron Technology (DIMES), Delft, The Netherlands,
for their support.
REFERENCES
[1] A. E. Ruehli, “Inductance calculations in a complex integrated circuit
environment,” IBM J. Res. Develop., vol. 16, pp. 470–481, 1972.
[2] H. M. Greenhouse, “Design of planar rectangular microelectronic induc-
tors,” IEEE Trans. Parts, Hybrids, Packag., vol. 10, pp. 101–109, 1974.
[3] R. L. Remke and G. A. Burdick, “Spiral inductors for hybrid and mi-
crowave application,” in IEEE Proc. 24th Electronic Computation Conf.,
1974, pp. 152–161.
[4] P. A. Brennan, N. Raver, and A. E. Ruehli, “Three-dimensional induc-
tance computations with partial element equivalent circuits,” IBM J. Res.
Develop., vol. 23, pp. 661–667, 1979.
[5] E. Pettenpaul and H. Kapusta, “CAD models of lumped elements on
GaAs up to 18 GHz,” IEEE Trans. Microwave Theory Tech., vol. 36, pp.
294–304, Feb. 1988.
[6] A. C. Cangellaris, J. L. Prince, and L. P. Vakanas, “Frequency-dependent
inductance and resistance calculation for three-dimensional structures in
high-speed interconnect systems,” IEEE Trans. Comp., Hybrids, Manu-
fact. Technol., vol. 13, pp. 154–159, Mar. 1990.
[7] N. M. Nguyen and R. G. Meyer, “Si IC-compatible inductors and LC
passive filters,” IEEE J. Solid-State Circuits, vol. 25, pp. 1028–1031,
Aug. 1990.
[8] R. G. Arnold and D. J. Pedder, “Microwave characterization of
microstrip lines and spiral inductors in MCM-D technology,” IEEE
Trans. Comp., Hybrids, Manufact. Technol., vol. 15, pp. 1038–1045,
Dec. 1992.
[9] R. B. Merril, T. W. Lee, H. You, R. Rasmussen, and L. A. Moberly, “Op-
Q
timization of high integrated inductors for multi-level metal CMOS,”
in Proc. Int. Electron Devices Meeting Tech. Dig., 1995, pp. 983–986.
[10] J. N. Burghartz, M. Soyuer, and K. A. Jenkins, “Microwave inductors
and capacitors in standard multilevel interconnect silicon technology,”
IEEE Trans. Microwave Theory Tech., vol. 44, pp. 100–104, Jan. 1996.
[11] C. P. Yue, J. Lau, T. H. Lee, and S. S. Wong, “A physical model for planar
spiral inductors on silicon,” in Proc. Int. Electron Devices Meeting Tech.
Dig., 1996, pp. 155–158.
[12] D. Lovelace, N. Camilleri, and G. Kannell, “Silicon MMIC inductor
modeling for high volume, low cost applications,” Microwave J., pp.
60–71, Aug. 1994.
[13] S. H. Song, H. B. Lee, H. K. Jung, S. Y. Hahn, K. S. Lee, C. Cheon, and
H. S. Kim, “Spectral domain analysis of the spiral inductor on multilayer
substrates,” IEEE Trans. Magn., vol. 33, pp. 1488–1491, Mar. 1997.
Authorized licensed use limited to: Tsinghua University. Downloaded on April 02,2023 at 11:57:47 UTC from IEEE Xplore. Restrictions apply.
1829
[14] J. R. Mosig and F. E. Gardiol, “A dynamic radiation model for mi-
crostrip structures,” in Advances in Electronics and Electron Physics,
P. W. Hawkes, Ed. New York: Academic, 1982, vol. 59, pp. 139–237.
[15] , “General integral equation formulation for microstrip antennas
and scatterers,” Proc. Inst. Elect. Eng., vol. 132, pp. 424–432, 1985.
[16] K. A. Michalski and D. Zheng, “Electromagnetic scattering and radia-
tion by surfaces of arbitrary shape in layered media. II. Theory,” IEEE
Trans. Antennas Propagat., vol. 38, pp. 335–344, Mar. 1990.
[17] M. Kahrizi, T. K. Sarkar, and Z. A. Maricevic, “Space domain approach
for the analysis of printed circuits,” IEEE Trans. Microwave Theory
Tech., vol. 42, pp. 450–457, Mar. 1994.
[18] R. Bunger and F. Arndt, “Efficient MPIE approach for the analysis of
three-dimensional microstrip structures in layered media,” IEEE Trans.
Microwave Theory Tech., vol. 45, pp. 1141–1153, Aug. 1997.
[19] T. Huan-Shang, L. J. Lin, R. C. Frye, K. L. Tai, M. Y. Lau, D. Kossives, F.
Hrycenko, and C. Young-Kai, “Investigation of current crowding effects
on spiral inductors,” in IEEE MTT-S Int. Microwave Symp. Dig., 1997,
pp. 139–142.
[20] A. M. Niknejad and R. G. Meyer, “Analysis, design, and optimization
of spiral inductors and transformers for Si RF ICs,” IEEE J. Solid-State
Circuits, vol. 33, pp. 1470–1481, Oct. 1998.
[21] , “Analysis of eddy-current losses over conductive substrates with
applications to monolithic inductors and transformers,” IEEE Trans. Mi-
crowave Theory Tech., vol. 49, pp. 166–176, Jan. 2001.
[22] B. Rejaei, J. L. Tauritz, and P. Snoeij, “A predictive model for Si-based
circular spiral inductors,” in IEEE MTT-S Int. Microwave Symp. Dig.,
1998, pp. 148–154.
[23] S. F. Mahmoud and E. Beyne, “Inductance and quality-factor evaluation
of planar lumped inductors in a multilayer configuration,” IEEE Trans.
Microwave Theory Tech., vol. 45, pp. 918–923, June 1997.
[24] J. N. Burghartz and B. Rejaei, “On the design of RF spiral inductors on
silicon,” IEEE Trans. Electron Devices, vol. 50, pp. 718–729, Mar. 2003.
[25] M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions
With Formulas, Graphs, and Mathematical Tables. New York: Wiley,
1972.
Behzad Rejaei received the M.Sc. degree in
electrical engineering from the Delft University of
Technology, Delft, The Netherlands, in 1990, and
the Ph.D. degree in theoretical condensed matter
physics from the University of Leiden, Leiden, The
Netherlands, in 1994.
From 1995 to 1997, he was a member of the
physics faculty with the Delft University of Tech-
nology, where he performed research on mesoscopic
charge–density–wave systems. Since 1997, he has
been with the Department of Electrical Engineering,
Mathematics, and Computer Science, Delft University of Technology, where
he is currently an Associate Professor. His research interests are in the areas of
EM modeling of integrated passive components and physics of ferromagnetic
devices.